Anything you forwarded to someone saying look at this.
Answer it out loud in the huddle, or fill in the full set at the brain-drain page.
Open — carried until closed
AskAn agent whose job is to interrogatecarried ×4
The strongest idea on the call. The system cannot know what you read, attended or were sent, so an agent should probe at each review for proprietary material worth ingesting.
Paul · opened 2026-08-10 · h4
RiskNobody has read a full report as a customer wouldcarried ×4
The machinery works; the writing has not been judged. A reading job, and it is Tim s to do.
Tim · opened 2026-08-10 · h6
RiskLinkedIn blocks bots - the gating technical problemcarried ×4
Requests 1 and 2 both depend on it. Everything else on the list is straightforward.
Paul · opened 2026-08-10 · h7
RiskAustin numbers have turnedcarried ×4
Metro wage growth flat and information-sector employment down two years running. Does not break the strategy, but grow with the boom is a weaker line than it was.
Paul · opened 2026-08-10 · h8
IdeaSpeaking slot at next years Ai4carried ×4
Roll out the marketing department. Paul view: should not take a year.
Tim and Melissa invited as optional. Graphed is in the same category as AME, but the intent is collaboration rather than competition -- they publish their method openly and do not gatekeep, and they are further along than we are. Worth deciding who should be on the call and what we would want from a working relationship.
Paul · opened 2026-08-10 · h11
AskWatch the rest of the Graphed video seriescarried ×3
Tim raised it as an action item: they have a whole series, not just this one. Do they go to that depth for every agent role? Paul: next on my list.
Paul · opened 2026-08-10 · h17
DecisionFinish ingesting their work BEFORE auditing our own agentscarried ×3
Tim proposed reversing the order and Paul agreed. The agent-by-agent audit still matters -- Tim called it the critical test -- but it comes after.
Paul · opened 2026-08-10 · h18
AskEvery agent needs role-level knowledge of how to do its jobcarried ×3
Tim s central point. An advertising agent does not natively know who to target. The LinkedIn method is the pattern; the question is how to do that for every role.
Paul · opened 2026-08-10 · h20
DecisionChannel order: LinkedIn first, X secondcarried ×3
Agreed on the call. Not Facebook or Instagram second -- Tim argued the people building this dialogue on X and have networks there.
Paul · opened 2026-08-10 · h21
AskDo the API calls outside Claude, not through itcarried ×3
Paul: do not let Claude kill you with API calls. Generate them natively instead. Paul to study and incorporate. Tim confirms this was a hot topic at Ai4 -- companies working out how to take work out of Claude.
Paul · opened 2026-08-10 · h22
AskTim and Melissa to check the huddle page dailycarried ×3
Paul asked both to bookmark wholereach.com/huddle and read it every morning. That is where everything now goes.
Tim · opened 2026-08-10 · h24
AskDawn to test the system and report backcarried ×3
Her role as stated on the call. Twenty five years from analog marketing through the ecommerce transition, mostly executive-assistant roles, two years working with AI professionally. Her observation: unless your thumb is on the pulse every heartbeat you fall behind, and keeping up alone is overwhelming. Tim agreed -- the pace is why nobody knows how to make an investment.
Dawn · opened 2026-08-10 · h26
RiskApify free credits exhausted - decide on paid tiercarried ×2
The 5 dollar monthly free allowance ran out partway through the home-building harvest: 19 of 60 builders completed, 41 returned actor-disabled. Magnetics finished fully. Apify paid tiers start around 39 dollars a month. Everything else works; this is purely a budget ceiling.
Paul · opened 2026-08-12 · h27
RiskModel API key needed for stages 2, copy and repliescarried ×2
ICP stage 2, campaign copy, reply classification and all organic drafting are blocked on one credential. Everything else is built and dry.
Paul · opened 2026-08-12 · h29
DecisionDo not polish UX until the functionality is settledcarried ×1
Tim and Melissa both, and they are right. The engine will change many times over the coming weeks; perfecting onboarding now means redoing it. Fix what is broken, defer what is merely rough. This is about respecting Dawn s time, not dismissing her findings.
Tim · opened 2026-08-13 · h48
DecisionTim and Melissa are not working on any magnetics sitecarried ×1
Confirmed on the call: no collision risk. PolyMagnet stays untouched so it can serve as a clean measure of the engine s net impact. They will tell us if that changes.
Tim · opened 2026-08-13 · h49
DecisionTim and Melissa joining the Cody call, 17 Augcarried ×1
Agreed. Tim sees him as a resource on whether full marketing automation is genuinely achievable - and if not, where it breaks.
Paul · opened 2026-08-13 · h50
ShippedLead capture is live on the windmill sitecarried ×1
The keystone of the campaign layer. /opt/wt-leads on :8956, proxied at royallswindmill.com/l/ so it runs on the client own domain - no CORS, no third-party script, nothing for an ad blocker to object to. Records taps on the phone, email and map links with full attribution (utm source/medium/campaign/content, else referrer host). Verified end to end in a browser: a tap arrived attributed to google/cpc/wells-test/local-angle. campaign.observed() now reads it, and reports call TAPS separately from LEADS - a tap is a proxy for a call, not a call, and merging them is the easiest lie a system like this can tell. Test rows cleared so day one starts clean.
Paul · opened 2026-08-13 · h51
AskDecide who answers a call-back form on a client sitecarried ×1
The form is written and NOT installed. A form creates an obligation: a rancher who fills it in at 9pm expects a call back. Committing the client to that is not ours to decide unilaterally. Click tracking needs no such decision - it measures behaviour that already happens - so that went live today and the form waits. If we install it, an instant ntfy push fires to wholetech-leads on every submission and someone has to own the response.
Paul · opened 2026-08-13 · h52
RiskMonday Cody call is on ZOOM - Tim and Melissa are on the wrong link
Two calendar entries existed for the 17 Aug 5:30pm slot. Cody organised the real one himself as a Graphed Discovery Call on Zoom (us02web.zoom.us/j/87135112318, ID 871 3511 2318). Pauls placeholder carried a Google Meet link and Tim and Melissa were on THAT one - so as it stood they would have sat in an empty Meet room. Placeholder retitled and repointed at the Zoom link; a leftover Join with Google Meet button may still show and should be ignored. Cody has been asked to add them to his invite properly. LIVE RISK until they are on his invite or have the Zoom link in hand.
Paul · opened 2026-08-14 · h53
ShippedSchneider Manual rewritten - 9 sections to 13
wholereach.com/ScheiderManual. New sections: the campaign layer (Tims spec, quoted and credited), lead capture and the tap-is-not-a-lead rule, knowledge-base-on-demand for any trade, and an honest split of everything still missing into what is a week of work and what is genuinely hard. Lessons updated with the quantified breadth finding (86/90 percent), the 55 percent employer-affinity finding, the 55 percent no-website finding, the significance ceiling, and the root cause of the wrong-industry bug. Costs corrected to 39 dollars a month actual. Footer carries a visible Graft to Graphed correction rather than a silent edit.
Paul · opened 2026-08-14 · h54
ShippedContacts added - Cody Schneider and Greg Isenberg
Cody: verified email from the calendar invite he sent, company Graphed, handles and the video citation. Greg: Late Checkout, handles only, email left BLANK deliberately and the note says so - a wrong address is worse than none. Also corrected the ingest notes: they recorded his company as Graft (graft.com), which is a parked domain listed for sale. The company is Graphed.
Paul · opened 2026-08-14 · h55
Closed since the last huddle
Turn what it finds into our own articles — Not just monitoring. Decide what to repost, and what should spur our own piece. Tim called it a self-fulfilling engine.
Cost control and ROI on the tool stack — Tim: if we add lots of costs the system price goes up and ROI gets harder to sell. A 100k system is a hard sell however good it is. Decide which paid tools earn their place.
Construction vertical is uncontested — Tim sizing: roughly 20,000 homebuilders, 100,000 trades, 25,000 manufacturers, all with websites, none doing automated marketing. His line: these are the people who will survive AI. Paul aim: a couple of hours a month to run their marketing department.
Home building beats magnetics, decisively — Same pipeline, one config file. Builders: 1,185 engagements from just 19 companies and 51 qualified - VP of Purchasing at KB Home, Purchasing Manager at Mungo Homes, Director of Production at LGI Homes, Design Studio Manager at Ashton Woods, CEO at Couto Homes. Magnetics: 1,074 engagements from all 43 companies and a list dominated by other suppliers sales staff. This answers h19.
Industry Intelligence served the wrong industry — Found by Dawn on her first day of testing. A well-drilling client was shown a competitive landscape of SEO software firms. No water-well KB existed, so the engine silently borrowed another trade s. It invented nothing - every company real, every description accurate, entirely the wrong industry. Harder to catch than fabrication because nothing looks wrong.
LinkedIn is the wrong source for local trades — Searching it for well drilling returned Schlumberger and Tenaris - offshore oil multinationals. A two-truck water-well outfit has no LinkedIn page at all. Google Maps is the right instrument: it classifies every business into a fixed category, so relevance is decided by Google rather than by keyword-matching a blurb. 90 found, 56 kept.
Half of a local trade has no website — 31 of 56 water-well businesses found in Texas have no working website - 23 list none at all and 8 list one that does not load. That is 55%, and it is not a gap in the data, it is the state of the trade. It is also the clearest sales argument the engine has: a business with no site cannot be marketed to by any of our competitors either.
First real ship to a client site — Dawn approved and shipped two schema additions to royallswindmill.com - a live client site we host and manage. Snapshot taken before each write, rollback available. First time the full loop - audit, activate, approve, ship - completed on a real site.
Run-after-approve race - fixed — The daily run loaded a workspace, generated for several seconds, then saved the stale copy - silently reverting any approval made in that window. Dawn lost one at 14:14 with no way to know. Now merges against disk instead of overwriting. Tested with a concurrent approve: survives.
Set-serialisation 500 - fixed — Activating a department returned a bare 500. Two scratch fields held Python sets, which cannot be JSON-encoded. Fixed at the serialiser so the whole class of bug cannot recur, and private scratch keys no longer leak into responses.
Pricing cut to 140 / 280 / 420 — Across 109 engine sites. The 2,900 agency-retainer comparison figure was deliberately left untouched - it is the anchor, not our price. The contrast is now 7x where it was 3x.
Skills library was stale on 86 sites, missing on 24 — The generator only ever wrote to automarketingengine.com. 111 sites request the library; 24 were 404ing, including automarketingagent.com. All now byte-identical, and the weekly refresh distributes automatically.
Build a KB on demand for any trade — New tooling: kb_from_maps.py and kb_from_linkedin.py turn a trade name into a KB config seed. Maps for local trades, LinkedIn for industries with corporate presence. This is how a client in a trade we have never covered gets real industry intelligence rather than someone else s.
A list of interested people is not automated marketing — Tim, 13 Aug, and it is the central critique. Google, Facebook and LinkedIn already let any business target people interested in a subject, natively, inside their own ad platforms. Finding the audience is step one of ten. Until the engine acts on that list it has nothing a customer would pay for.
Generate the content and the ad creative — Tim: the engine must produce articles, posts, landing pages, and ads in multiple formats - not briefs about them. Multiple value propositions, multiple creatives, several channels.
Test creatives and report cost per lead — Tim: run the variants, measure which message and which image actually pull, and come back with a real cost of acquisition per channel. His example: is it 25 percent off, or is it guaranteed service in 48 hours? Nobody knows until it is tested.
Propose experiments and ask permission to spend — Tim: the agent should say - I want to spend 250 dollars in Google on this creative to find your acquisition cost - then report back that it is 32 dollars a lead. The owner sets the threshold. This is the interaction model he wants.
Honour a budget ceiling and a cost-per-lead cap — Tim: two separate limits. I will pay 5 dollars a lead, AND do not bill me more than 1000 a month. Both must be respected, and both can move by season.
Exhaust the free channels before spending anything — Tim: given a 10k budget the system must not simply start buying ads. Articles, content, video, organic - all free - should always be running first and continuously.
Onboarding should recommend, not interrogate — Dawn: asking a non-marketer to choose goals and channels is asking them the question they hired you to answer. Take the business profile, run it through the engine, and recommend. Her words - put AI to work for me there.
Start from no budget, not from a big one — Dawn: assume we have five dollars and we are starving - how do we get a customer? Ask for 10k and someone will suggest a billboard. Free-first falls out of this naturally, and it matches Tim on the same point.
Navigation loses your place, and logs you out — Dawn on first use: clicking deeper is easy, retracing is not. The back arrow logs you out. Noted but deliberately NOT scheduled - see the UX-timing decision.